TIdeS

Transcript Identification and Selection (TIdeS) is a method to identify putative open reading frames (pORFs) from a given transcriptome and is able to aid in the bulk decontamination of sequences from "messy" transcriptomic data.

Overall, TIdeS couples sequence composition with ML approaches to discern pORFs in the correct reading frame with substantial improvement over other popular tools, while providing support for additional non-standard genetic codes. Additionally, TIdeS can be used to classify ORFs into several user-defined categories from highly contaminated datasets (e.g., parasite + host, kleptoplasts, big "dirty" protists).

Dependencies

• Python 3.7+
  • BioPython 1.79+
  • Pandas 2.0.3+
  • scikit-learn 1.3+
  • Optuna 3.3.0+
  • seaborn 0.12.2+
• DIAMOND 2.0.13+
• CD-HIT 4.8.1+
• Barrnap 0.9
• Kraken2 2.0.0+

Installation

Note that TIdeS is only supported on UNIX systems (linux and MacOS).

Python's pip can be used to install the necessary python version and related packages.

pip install tides-ml

Followed by downloading the precompiled executables for the remaining dependencies.

Alternatively, you can do this through conda (note this will be updated):

# Create a new environment for TIdeS
conda create -n tides-ml
conda activate tides-ml

# Install the necessary packages (with minimum support versions)
conda install -c bioconda -c conda-forge diamond">=2.0.13" cd-hit">=4.8.1" barrnap">=0.9" kraken2">=2.0"

Clone the repository.

git clone https://github.com/xxmalcala/TIdeS.git

Running TIdeS

The general syntax to run TIdeS is:

tides --fin <transcriptome-assembly> --taxon <taxon-name> --db <protein-database>

Several example command lines and uses for TIdeS (i.e., ORF-calling and ORF classifying) are included in the examples folder. To run the examples, you need to be within the examples folder (e.g., ./orf_call_and_decontam.sh)

List of all options

ORF-Calling and Assessment

Reference protein database

Create a reference protein database for TIdeS (note you can use your own if you choose!). This will generate a database from six diverse eukaryotes, representing a broad yet compact database for subsequent ORF-calling.

Note that this database (tides_aa_db.dmnd) will be prepared from whichever directory you call upon this script.

./TIdes/util/prep_tides_db.sh

Inputs

• FASTA formatted transcriptome assembly
• Taxon name (e.g., Homo sapiens, Op_me_Hsap)
• Protein database (can be prepared by "prep_tides_db.sh" in the util folder)

tides -i <transcriptome-assembly> -o <taxon-name> -d <protein-database>

Decontamination

Inputs

• FASTA formatted transcriptome assembly
• Taxon/project name (e.g., Durisnkia baltica, Dinotoms) Optional (need one)
• Table of annotated sequence names (see examples folder)
• Path to Kraken2 database to identify putative non-eukaryotic sequences to remove

With a table of annotated sequences:

tides -i <transcriptome-assembly> -o <taxon-name> -c <annotated-seqs-table>

Using a Kraken2 database:

tides -i <transcriptome-assembly> -o <taxon-name> -c -k <kraken2-database>

Table of annotated sequences

The <annotated-seqs-table> should include sequence names and their category separated by tabs. Note that these sequences should be present within the input FASTA file as well. Please aim to include at least 25 sequences for each category, although more (up to ~200) is great!

seq1  human
seq2  lunch
seq3  lunch
seq4  human
seq5  lunch
...

Additional uses/approaches

More on how to run TIdeS and its uses can be found in the examples folder, including:

• ORF-Calling
• Classification of ORFs
• ORF-calling and sequence classifying with a previously trained model
• Preparing a simple proteome database and ORF-calling
• Naive approaches to inferring contamination
• Example FASTA and <annotated-seqs-table> files